Settings
Settings opens as a side sheet from the sidebar. The panels are General, Model, Transcribe, AI, Advanced and License.
General
- Language: nine UI languages.
- Theme: follow the system, light or dark.
- Mirror: download mirror for models and helper binaries.
- Analytics: anonymous usage analytics, on by default; can be turned off.
- Storage and backup: database backup and restore, log export.
Model
Model Path
Directory where local ASR models are stored — Whisper, Qwen3-ASR and SenseVoice. Models download to the default path on first use; you can point the folder elsewhere.
Engines and Models
- Whisper:
tiny→large-v3-turbo; VAD filtering and stable timestamps. Fastest to most accurate. - Qwen3-ASR:
0.6B(~1.7 GB) and1.7B(~4.5 GB); strongest for Chinese, dialects and Chinese-English code-switching. Add the optional Forced Aligner for frame-accurate subtitle timing. - SenseVoice: sherpa-onnx, runs in-process; available on macOS and Linux today (Windows is not verified yet).
Switch the engine in Settings → Model. See Qwen3-ASR Engine and Transcription.
GPU Acceleration
- macOS: CoreML / ANE on Apple Silicon
- Windows / Linux: choose the GPU device index; Vulkan is supported
Transcribe
Language
auto: detect automatically (slightly lower accuracy)- A fixed language, e.g.
zh,en,ja
Temperature
Controls sampling randomness. 0 is deterministic, 1 is the most random. Default 0; keep it unless you have a reason.
Stable Timestamps
Silero VAD-backed precise timestamp mode. More accurate subtitle timing when enabled, slightly slower.
Speaker Diarization
Downloads an extra model on first use. Up to four speakers.
Audio Enhancement
Denoise, voice boost and telephone-audio presets for low-quality recordings.
Recording
- Default device: microphone and system-audio device
- Format: WAV (lossless) or MP3 (compressed)
- Save path: default location for recordings
- Auto-transcribe: start transcription after recording
AI
LLM Provider
- Local: Ollama
- Cloud: Claude, OpenAI-compatible, DeepSeek, Gemini, Zhipu, DashScope (Qwen), OpenRouter, MiniMax
API Key and Endpoint
Enter the key and, when needed, the base URL. Keys are stored locally and never uploaded.
Prompt Templates
Edit the default prompts for summary, insights, chat, chapters, translation, proofread, cleanup, line breaking and dictation cleanup. Prompts bound to a parser (proofread, line breaking, translation) are shown read-only.
Export
- Default format: default export format
- Include timestamps: whether exports carry time information
- Encoding: UTF-8 or UTF-8 with BOM
Advanced
Watch folders, Webhook, remote locations (WebDAV / SMB), backups, shortcuts and log export.
Shortcuts
| Shortcut | Function |
|---|---|
⌘K / Ctrl+K | Command palette |
⌘O / Ctrl+O | Open file |
⌘[ / Ctrl+[ | Back to library |
⌘F / Ctrl+F | Search the library |
⌘, / Ctrl+, | Settings |
⌘= / ⌘- / ⌘0 | UI zoom |
⇧⌘W / Ctrl+Shift+W | Close media |
Space / K | Play / pause |
← / → | Seek 5 s (Shift: 1 s) |
, / . | Slower / faster |
L / [ / ] | Loop, set loop points |
M | Mute |
C | Subtitles |
E | Edit the current cue |
F | Fullscreen |
S | Screenshot |
I | Toggle the inspector |
B | Highlight the current line |
? | Shortcut help |
⌘⏎ / Ctrl+Enter | Split the current line |
⌘J / Ctrl+J | Merge the next line |
⌘Z / ⇧⌘Z | Undo / redo |